Abstract

This dissertation presents and discusses a methodology for automated decision problem formulation. More precisely, it describes the design of a computer system that can derive an analytic model of a decision problem through an interactive dialog with the decision maker. This design task would typically require prior knowledge of the decision problem domain, or assume decision-analytic expertise of the decision maker to guide the formulation process. The central thesis of this dissertation, however, is the claim that we can automate the formulation of an arbitrary decision problem without either one of these two assumptions. ;This claim addresses both a technical and a human domain. While the demands of the technical domain can be described and satisfied in the context of traditional, scientific arguments, the human challenges involved cannot be addressed in the same terms. Phenomena of perception, understanding and judgment occur during the decision problem formulation dialog that are invisible in the technical domain, and that cannot be explained in terms of our traditional, scientific school of thought. This school of thought is largely based on the philosophy of Descartes. It is rationalistic, in that it accepts rational thought as the only valid source of argument. To complement Descartes' philosophy, this dissertation presents the nonrationalistic viewpoints of Heidegger, Wittgenstein, Merleau-Ponty and others, and reveals their strong relevance to decision analysis in general and to automated decision problem formulation in particular. ;On the background of a strong distinction between rationalistic and nonrationalistic philosophies, the proposed methodology is described from a phenomenological, conceptual and technical perspective, which differ fundamentally in their levels of abstraction. The technical view, in particular, supports the test implementation and real-world application of the methodology to personal, humanitarian and business decisions, which is also documented in this dissertation